perf: parallelize title and tag generation with ThreadPoolExecutor

Submit generate_title and generate_tags concurrently via ThreadPoolExecutor(max_workers=2) in _process_audio, eliminating 2-4s of sequential LLM latency. Adds a concurrency test to verify both calls start within 20ms of each other.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: 2 people

tests/test_main.py

@@ -151,3 +151,77 @@ def test_setup_argparser_has_reset_personas_flag():
             main()
 
     mock_seed.assert_called_once_with(force=True)
+
+
+def test_title_and_tags_generated_in_parallel(tmp_path):
+    """generate_title and generate_tags should start concurrently."""
+    import time
+    from datetime import datetime
+    from unittest.mock import MagicMock, patch
+    from tinysteno.main import _process_audio
+    from tinysteno.personas import Persona
+    from pathlib import Path
+
+    persona = Persona(
+        slug="default",
+        name="Default",
+        description="desc",
+        schema={"summary": {"type": "string", "description": "summary"}},
+        system_prompt="You are a test assistant.",
+        template="{{ title }}",
+        template_path=Path("/fake/template.md"),
+    )
+    config = {
+        "api_key": "ollama",
+        "base_url": "http://localhost",
+        "model": "test",
+        "whisper_model": "small",
+        "diarization": False,
+        "auto_title": True,
+        "auto_tags": True,
+        "obsidian_vault": str(tmp_path),
+        "output_folder": "meetings",
+    }
+
+    call_start_times: dict = {}
+
+    def fake_title(text):
+        call_start_times["title"] = time.monotonic()
+        time.sleep(0.05)
+        return "Test Title"
+
+    def fake_tags(text):
+        call_start_times["tags"] = time.monotonic()
+        time.sleep(0.05)
+        return ["test"]
+
+    mock_transcribe_result = {
+        "text": "hello world",
+        "diarised_text": "",
+        "detected_language": "en",
+        "duration_seconds": 10.0,
+    }
+
+    with patch("tinysteno.main.WhisperTranscriber") as mock_tc, \
+         patch("tinysteno.main.Orchestrator") as mock_oc, \
+         patch("tinysteno.main.ObsidianExporter") as mock_ec:
+
+        mock_tc.return_value.transcribe.return_value = mock_transcribe_result
+        mock_orch = mock_oc.return_value
+        mock_orch.summarize.return_value = {"summary": "hello world"}
+        mock_orch.generate_title.side_effect = fake_title
+        mock_orch.generate_tags.side_effect = fake_tags
+        mock_ec.return_value.export.return_value = tmp_path / "note.md"
+
+        _process_audio(
+            wav_path=str(tmp_path / "audio.wav"),
+            name=None,
+            config=config,
+            logger=MagicMock(),
+            persona=persona,
+            timestamp=datetime(2024, 1, 1),
+        )
+
+    assert "title" in call_start_times and "tags" in call_start_times
+    overlap = abs(call_start_times["title"] - call_start_times["tags"])
+    assert overlap < 0.02, f"title and tags should start concurrently, gap={overlap:.3f}s"

tinysteno/main.py

@@ -157,21 +157,31 @@ def _process_audio(  # pylint: disable=too-many-arguments,too-many-positional-ar
                 all_items.extend(data.get(field, []))
         first_string_value = ". ".join(all_items) if all_items else None
 
+    # Generate title and tags in parallel when both are enabled
+    from concurrent.futures import ThreadPoolExecutor
+
+    title_future = None
+    tags_future = None
+
+    if (config.get("auto_title") or config.get("auto_tags")) and orchestrator and first_string_value:
+        print("Generating title and tags...")
+        with ThreadPoolExecutor(max_workers=2) as executor:
+            if config.get("auto_title"):
+                title_future = executor.submit(orchestrator.generate_title, first_string_value)
+            if config.get("auto_tags"):
+                tags_future = executor.submit(orchestrator.generate_tags, first_string_value)
+
     # Resolve title
-    title = name  # start with --name if provided
+    title = name
     if not title:
-        if config.get("auto_title") and orchestrator and first_string_value:
-            print("Generating title...")
-            generated = orchestrator.generate_title(first_string_value)
+        if title_future is not None:
+            generated = title_future.result()
             title = generated if generated else Path(wav_path).stem
         else:
             title = Path(wav_path).stem
 
     # Resolve generated tags
-    generated_tags: list = []
-    if config.get("auto_tags") and orchestrator and first_string_value:
-        print("Generating tags...")
-        generated_tags = orchestrator.generate_tags(first_string_value)
+    generated_tags: list = tags_future.result() if tags_future is not None else []
 
     date_str = timestamp.strftime("%Y-%m-%d %H:%M")
     duration_str = _format_duration(result.get("duration_seconds", 0.0))